Classification of fish diseases stands as a vital underwater aquaculture task which enables prompt detection and control of infections. Convolutional neural networks (CNNs) enable deep learning to act as a strong automated disease detection mechanism through which complex image patterns get extracted. The research develops a YOLOv9 based deep learning model enhancement which performs classification on 11 fish diseases. The improved version of this model includes extra convolutional layers with connected layers supporting a classification layer through Softmax activation for better fish disease detection performance. Evaluation and training tasks were performed on FishLens dataset. Our proposed model reaches an mAP@50 evaluation value of 81% while surpassing YOLOv8 detection by 2%. Real time fish disease classification effectiveness emerges from our method which provides an automated monitoring solution for aquaculture systems.

Accurate prediction of liver diseases such as fatty liver, cirrhosis, and liver cancer remains a critical challenge in modern healthcare due to increasing prevalence and limitations in existing diagnostic methods. Current approaches often rely on single data modalities, lack interpretability, and compromise data privacy through centralized models, limiting their scalability and applicability in real-world clinical settings, especially in sparse and noisy data environments. This research aims to address these challenges through a multi-objective framework encompassing advanced machine-learning techniques. The first objective focuses on developing a clinical imaging-genetic data fusion model using self-supervised graph neural networks to manage sparse datasets, achieving accuracy gains of 92-94% and improving AUC-ROC scores by 5-8% over baseline models. The second objective introduces the Federated Meta-Learning Framework for Liver Disease Prediction (FML-LDP), enabling privacy-preserving, collaborative model training across institutions. This framework achieves precision levels of 88-91%, reduces computational overhead by 15-20%, and ensures adaptability to diverse patient scenarios. The third objective addresses the need for interpretability through the Explainable Deep Learning Framework with Reinforcement Learning Optimization (EDL-RL), which dynamically selects optimal features using reinforcement learning. This framework enhances interpretability by 25-30%, integrates Shapley-value-based feature explanations, and maintains high predictive accuracy (90-93%) to improve clinical trust.By integrating multi-modal data learning, privacy-preserving collaboration, and interpretable AI, this work provides a robust and scalable solution for liver disease prediction, setting a new benchmark for clinical decision support systems.

Accurate pneumonia diagnosis is crucial for reducing mortality rates, particularly in resource-constrained healthcare facilities. A deep learning detection framework that examines pneumonia diagnosis for chest X-ray images constitutes the proposal of this research. This system leverages the EfficientNetB7 architecture with Squeeze-and-Excitation (SE) blocks, which significantly increases feature extraction and outperforms the baseline EfficientNetB0 in distinguishing pneumonia from normal cases. The data undergoes systematic division into three parts for training, validation, and testing purposes as part of a thorough model evaluation. The final model achieves an impressive detection accuracy of 98.31%, surpassing existing approaches in this domain. To enhance interpretability, Grad-CAM heat maps are employed to highlight the most influential regions in the X-ray images, aligning with clinical diagnostic needs. This visualization-driven approach improves trust and transparency in AI-assisted medical decision-making, making it a valuable tool for pneumonia diagnosis.

To decipher human activities and facilitate organic computer-human interactions, facial expression recognition is crucial using four datasets JAFFE, CKPlus, KDEF and AffectNet, we achieve excellent accuracy in human face emotion classification with the VGG 16 pre-trained model with transfer learning. To obtain a comprehensive insight of the model's decision-making process, we employ Layerwise Relevance Propagation (LRP), a method from explainable Artificial Intelligence (XAI). Only positive relevance scores are taken into account for successfully predicted test images from the datasets. Contributory pixels towards predicting the intensity of emotion are pixels with good relevance ratings. By combining emotion recognition with LRP, we can forecast emotion labels and ranks. Using a confusion matrix, we checked if our predictions were in line with reality. Our model achieved intensity prediction accuracy of 96.33% on JAFFE, 95.78% on CKPlus, 95.78% on KDEF, and 93.89% on AffectNet. A group of ten annotators work together to generate ground truth by assigning ratings of "MINIMAL, "AVERAGE, and "STRONG to each image. This study demonstrates how well our method predicts the ranks of emotion intensity and provides information on how trustworthy and interpretable the model is. Facial emotion recognition's intensity ranking is made more robust with the addition of XAI techniques like LRP.

Fish is a key protein source, with rising global demand placing pressure on aquaculture to maintain fish health and productivity. However, diseases like Bacterial Gill Disease, Aeromoniasis, and Parasitic Disease pose serious threats to the industry, and early detection is crucial to prevent outbreaks. This study investigates the application of deep learning models, particularly EfficientNet and CBAM (Convolutional Block Attention Module), to classify fish diseases with improved accuracy. Our proposed models were tested on a multi-class fish disease dataset, where EfficientNetB6+CBAM achieved the highest classification accuracy of 99.45% and a superior F1-score across disease types. The results emphasize the significance of attention mechanisms in enhancing feature extraction from complex datasets, providing a highly accurate and accessible disease detection tool for fish farmers. This approach marks a promising advancement in sustainable aquaculture by facilitating intelligent disease management through deep learning technologies.

As for the contribution of this research, an Automatic Nutrient Prediction System (APS) is introduced, which applies deep learning with CNN to recognize food images for nutritional advice. Using the large Food-101 dataset of 101,000 images split into 101 food classes, the proposed CNN model comprises five Convolutional Layers Conv2D, MaxPooling; and Subsampling Layers (values of 0.25 and 0.4) to combat overfitting. Training data is normalized across RGB channels with the sizes of the images decreasing to 128 x 128 pixels. The data is augmented through rotation, translation, and brightness changes. Training accuracies of 98.76 and 98.87 across epochs demonstrate the model’s potential for future improvements including Vision Transformers with RGB-D, for sharp volumetric accuracy and inference. The applications include foods that will be recommended for an individual’s proper diet based on this system in the health and wellness industry.

Facial emotion recognition (FER) plays a vital role in human-computer interaction, enabling applications starting from sentiment analysis to affective computing. In this study, we propose an innovative Multi-stage Hierarchical Attention Network model for Facial Emotion Recognition in Wild (MHAN-FERW). This model is designed to address the challenges encountered in uncontrolled environmental conditions when it comes to recognizing facial expressions. The MHAN-FERW architecture utilizes an EfficientNetB6 backbone network to extract features, followed by the application of an attention mechanism to capture spatial, channel-wise attention and temporal information, added with a feature pyramid network and fully connected layers to enhance robust feature representation. Through experimentation on the AffectNet dataset, MHAN-FERW achieves a notable accuracy of 67%, surpassing the state-of-the-art benchmarks by 3%. The proposed MHAN-FERW architecture demonstrates the efficacy of integrating attention mechanisms and feature pyramid networks for facial emotion recognition in challenging real-world scenarios. The insights gained from this study contributed in the advancement of FER which has potential applications in various fields of computing.

The field of Facial Emotion Recognition (FER) has advanced considerably in the last few years. Much research relies on lab-controlled datasets, characterized by limitations in size, quantity, and quality. These datasets feature high-resolution static images captured in ideal conditions but lack fidelity in representing real-world scenarios. Hence, FER systems must be trained on primary data that includes real-world scenarios like facial expressions captured from various angles and in different lighting conditions, images with occlusion etc., broadly termed as unconstrained environment. To leverage the gap, this study emphasizes utilizing an AffectNet dataset that has samples close to real-world scenarios. In addition, we propose a novel ensemble framework to increase the accuracy of emotion recognition by harnessing the complementary strengths of three distinct deep-learning models: DenseNet169, EfficientNetB7 and InceptionV3. The key innovation lies in our novel ranking-based fusion technique, which introduces a unique perspective on model confidence and its relationship with prediction quality. The rank-based fusion approach optimally harnesses each base model's unique characteristics and strengths. Our experiments confirm the ensemble framework's effectiveness, outperforming individual models in facial emotion recognition.

As a result of their surroundings and lifestyle choices, people nowadays suffer from a wide range of ailments. As a result, predicting illness at an early stage is crucial. Doctors, on the other hand, struggle to make accurate diagnoses based solely on symptoms. The most challenging task is predicting sickness properly. Machine learning plays a key part in forecasting in order to complete this difficult task. To tackle this challenge, machine learning plays a key role in illness prediction. Medical research creates a vast amount of data every year. Early patient care has benefitted from effective medical data analysis because of the rising quantity of data growth in the medical and healthcare professions. In data mining, disease data is utilised to identify hidden patterns in huge volumes of medical data. Based on the patient's symptoms, we created a broad disease prediction. Machine learning algorithms like ANFIS and CNN are used to properly predict sickness (adaptive network-based fuzzy inference system). The collection of illness symptoms is necessary for disease prediction. For an accurate prognosis, this general illness prediction takes into account the person's lifestyle and medical history. When it comes to illness prediction, ANFIS outperforms CNN by a wide margin (96.7%). ANFIS, on the other hand, does not require as much time or memory to train and test because it does not use the UCI repository dataset. There are several libraries and header files included with the Anaconda (Jupyter) notebook that make Python programming more precise and accurate.

Researchers are interested in Facial Emotion Recognition (FER) because it could be useful in many ways and has promising applications. The main task of FER is to identify and recognize the original facial expressions of users from digital inputs. Feature extraction and emotion recognition make up the majority of the traditional FER. Deep Neural Networks, specifically Convolutional Neural Network (CNN), are popular and highly used in FER due to their inherent image feature extraction process. This work presents a novel method dubbed as EfficientNet-XGBoost that is based on Transfer Learning (TL) technique. EfficientNet-XGBoost is basically a cascading of the EfficientNet and the XGBoost techniques along with certain enhancements by experimentation that reflects the novelty of the work. To ensure faster learning of the network and to overcome the vanishing gradient problem, our model incorporates fully connected layers of global average pooling, dropout and dense. EfficientNet is fine-tuned by replacing the upper dense layer(s) and cascading the XGBoost classifier making it suitable for FER. Feature map visualization is carried out that reveals the reduction in the size of feature vectors. The proposed method is well-validated on benchmark datasets such as CK+, KDEF, JAFFE, and FER2013. To overcome the issue of data imbalance, in some of the datasets namely CK+ and FER2013, we augmented data artificially through geometric transformation techniques. The proposed method is implemented individually on these datasets and corresponding results are recorded for performance analysis. The performance is computed with the help of several metrics like precision, recall and F1 measure. Comparative analysis with competent schemes are carried out on the same sample data sets separately. Irrespective of the nature of the datasets, the proposed scheme outperforms the rest with overall rates of accuracy being 100%, 98% and 98% for the first three datasets respectively. However, for the FER2013 datasets, efficiency is less promisingly observed in support of the proposed work.

The internet of things (IoT) is a revolutionary communication paradigm that aims to offer manifold new services in the context of aquaculture. Advancement of embedded chips and computers like Audrion, raspberry pi has provided a great platform to leverage the services in both agriculture and aquaculture. The quality of water parameters plays a critical role in rising aqua organisms like fish and prawns. The data from the sensors like pH, temperature, TDS, etc., are transmitted to the cloud as well as to the farmer. In this chapter, we propose an IoT-based monitoring mechanism for the water quality parameters, a vital parameter for ensuring healthy aquaculture and hence increase productivity. By using PyCharm, we train a model that suggests for a solution on encountering a problem. A new method is proposed for feeding the organisms with the help of a feed sensor. Feed sensor senses the behavior of the hungry fishes and signals the feeder to deliver the foodstuffs to the pond. This reduces majority of the food waste and hence saves money. In this chapter, the quality of water parameters is trained using support vector machine (SVM).

With the change from laboratory controlled to challenging facial expression recognition (FER) in the wild and the recent success of deep learning techniques in different fields, deep neural networks have been increasingly leveraged for automated FER to learn discriminatory representations. Here, in this survey, we include a brief overview of deep FER literatures and provide insights into some essential issues. Firstly, we represent the existing datasets that are widely used for the purpose and then we define a deep FER system’s standard pipeline with the associated context information and suggestions for applicable executions for each level. We then present already existing novel deep neural networks (DNN) and related training approaches for the state-of-the-art deep FER techniques that are optimized on the basis of both static and dynamic image sequences. A competitive comparison of the experimental works is also presented along with an analysis of relevant problems and implementation scenarios. Lastly, an overview of the obstacles and appropriate opportunities in this area is presented.

Given the significance of speed in the realm of the Internet and the large number of cyberattacks, verification systems that are fast, accurate, and convenient are required. Although it is possible to manipulate Image Recognition verification, it can still be of some help against any form of fraudulent scheme. The present study proposes a model of pixel-wise operations for identifying a location point. The computer vision is not limited to pixel-wise operations. It can be more complicated than image processing. First, unstructured image segmentation is taken via K-Means Clustering Algorithm. Then, after completing the preprocessing step, the segmented image is extracted from the surveillance cameras to identify expressions and vehicle images. Raw images of the surveillance camera comprise the images of individuals and vehicles that are classified by means of DWT. Further, the images that represent the appearances are taken by Smart Selfie Click (SSC). These two features are extracted in order to identify whether or not a vehicle should be permitted into the campus, thus making the verification possible. These two images are nothing but reliable objects extracted and used for location identification.

In ongoing year, the expanding number of wearable sensors on human can fill for some, needs like crisis care, medicinal services remote observing, individual diversion and correspondence and so forth. The WBAN empowers the clinical applications to be created utilizing electronic gadgets and sensors. The WBAN is made by wearing little sensors on the human body. A significant issue is the means by which to protect security and protection of patient's clinical social insurance data over remote correspondence. The vitality utilization and information security are as yet significant difficulties in medicinal services applications. In this paper we proposed an algorithm to secure patients data from eavesdropping attack with help of RSA secret key encryption technique. The target of this paper is to propose a remote sensor organize framework in which both pulse and internal heat level of various patients can screen on PC simultaneously utilizing RF arrange. The proposed model framework incorporates two sensor hubs and beneficiary hub (base station). The sensor hubs can transmit information to collector utilizing remote nRF handset module. The nRF handset module is utilized to move the information from microcontroller to PC and a graphical UI (GUI) is created to show the deliberate information and spare to database. Numerous advances have demonstrated their productivity in supporting WBANs applications, for example, remote observing, biofeedback and helped living by reacting to their particular nature of administration (QoS) necessities. Because of various accessible innovations, choosing the fitting innovation for a clinical application is being a difficult undertaking. In this paper, the distinctive clinical applications are introduced. The most widely recognized advances utilized in WBANs are featured. At long last, a coordinating between every application and the relating reasonable innovation is considered.